Cisco UCS FI 6248 connectivity with Cisco catalyst 2960

[rahul.maheshwari1]

In our UCS environment, both fabric interconnects has Cisco Nexus 9k as upstream switches with vPC and it is running fine. But we need to isolate some virtual servers on UCS blades on a entirely separate DMZ switch which is Cisco catalyst 2960.

1) So can we connect separate physical twinax cables from  FI uplink ports to catalyst 2960 and do the connectivity for DMZ servers by keeping the FI to nexus connectivity as it is ?

2) In this case, as there are 2 core nexus switches 1 and 2 so will we require 2 cisco catalyst 2960 for this disjointed kind of network ? or else we can connect FI A and FI B to a single 2960 switch on its 2 nos. of 10 Gig SFP+ ports ? 

3) also suggest things needs to be taken care, best practise or illustration guides in this context .  

[Walter Dey]

Hi Rahul

You will do a "disjoint vlan" design, see eg.

http://www.cisco.com/c/en/us/solutions/collateral/data-center-virtualization/unified-computing/white_paper_c11-692008.html

This paper will answer all your questions.

Walter.

[rahul.maheshwari1]

Thanks for sharing this.  We will work out according to this and believe that we will achieve this for north bound setup. But, I have some queries in south bound setup. 

At present, we are using only 2 out 4 IOM module ports on Chassis nos. 1 for production setup. There are some virtual machines in blade nos. 4 / Chassis 1 which needs to be in DMZ network.  For DMZ network, we want to have dedicated links between IOM and FI for a blade server. This is what we have planned :- 

• 1) We will use 2 more links on port no. 3 and 4 of IOM module to FI. So, now we will be utilizing all 4 IOM links in 2204 IOM module. We will reacknowledge this Chassis no. 1.

Here, my query is that Is it possible to keep 2 links in port channel admin state between IOM and FI for Chassis no. 1 . And remaining 2 IOM links in ‘None’ admin state so that these remaining 2 IOM links can be used for blade which will be in DMZ. 

[Kirk J]

Greetings.

The Blade adapter to IOM assignment, and chassis IOM connectivity is automated.

You cannot manually assign certain blades to certain IOM uplinks.  Your chassis discovery policy will either set all of your links to individual or in port-channel grouping (6200 series and above).

Thanks,

Kirk...

[Walter Dey]

Slight correction:

you have 8 slots per Chassis, and between 1 and 8 IOM-FI uplinks.

If you have 8 uplinks (no pc) then you achieve a dedicated IOM uplink per slot/blade. It's of course limited to 10g per fabric.

e.g. slot 1 - uplink 1, slot 2 - uplink 2,.......

If you have fewer uplinks, you share the bandwith between slots / blades.

[rahul.maheshwari1]

Thanks Walter, Got your point now.  In your example case, if all 8 uplinks are configured as discrete (no port channel) , then we achieve dedicated IOM uplinks per slot/blade. But these IOM uplinks will remain dedicated to per slot/blade or else their arrangement is subjected to change randomly to any slot/blade.  In other words, once we configure in this way, will we be able to manually assign an IOM link to a particular blade server ? If yes, then we configure a particular IOM link to a particular blade server, will it remain constant or it will change randomly ?  

[Walter Dey · Accepted solution]

The assignment is static and cannot be changed !

slot 1 - uplink 1

slot 2 - uplink 2,.......

If a slot has no blade, the corresponding uplink is not used and this cannot be changed !

This dedication of IOM uplinks is of course resource intensive: cables, ports on FI, port license,.....